Mining Significant Maximum Cardinalities in Knowledge Bases

International audienceSemantic Web connects huge knowledge bases whose content has been generated from collaborative platforms and by integration of heterogeneous databases. Naturally, these knowledge bases are incomplete and contain erroneous data. Knowing their data quality is an essential long-term goal to guarantee that querying them returns reliable results. Having cardinality constraints for roles would be a significant advance to distinguish correctly and completely described individuals from those having data either incorrect or insufficiently informed. In this paper, we propose a method for automatically discovering from the knowledge base’s content the maximum cardinality of roles for each concept, when it exists. This method is robust thanks to the use of Hoeffding’s inequality. We also design an algorithm, named C3M, for an exhaustive search of such constraints in a knowledge base benefiting from pruning properties that drastically reduce the search space. Experiments conducted on DBpedia demonstrate the scaling up of C3M, and also highlight the robustness of our method, with a precision larger than 95%